A computer network typically comprises a plurality of interconnected entities that transmit (i.e., “source”) or receive (i.e., “sink”) data frames. A common type of computer network is a local area network (“LAN”) which typically refers to a privately owned network within a single building or campus. LANs employ a data communication protocol (LAN standard), such as Ethernet, FDDI or token ring, that defines the functions performed by the data link and physical layers of a communications architecture (i.e., a protocol stack), such as the Open Systems Interconnection (OSI) Reference Model. In many instances, multiple LANs may be interconnected by point-to-point links, microwave transceivers, satellite hook-ups, etc. to form a wide area network (“WAN”), metropolitan area network (“MAN”) or intranet. These LANs and/or WANs, moreover, may be coupled through one or more gateways to the Internet.
One or more intermediate devices are often used to couple LANs together and allow the corresponding entities to exchange information. For example, a bridge may be used to provide a “bridging” function between two or more LANs. Alternatively, a switch may be utilized to provide a “switching” function for transferring information, such as data frames, among entities of a computer network. Typically, the switch is a computer having a plurality of ports that couple the switch to several LANs and to other switches. The switching function includes receiving data frames at a source port and transferring them to at least one destination port for receipt by another entity.
Switches may operate at various levels of the communication protocol stack. For example, a switch may operate at layer 2 which, in the OSI Reference Model, is called the data link layer and includes the Logical Link Control (LLC) and Media Access Control (MAC) sub-layers. Data frames at the data link layer typically include a header containing the MAC address of the entity sourcing the message, referred to as the source address, and the MAC address of the entity to whom the message is being sent, referred to as the destination address. To perform the switching function, layer 2 switches examine the MAC destination address of each data frame received on a source port. The frame is then switched onto the destination port(s) associated with that MAC destination address.
Other devices, commonly referred to as routers, may operate at higher communication layers, such as layer 3 of the OSI Reference Model, which in TCP/IP networks corresponds to the Internet Protocol (IP) layer. Data frames at the IP layer also include a header which contains an IP source address and an IP destination address. Routers or layer 3 switches may re-assemble or convert received data frames from one LAN standard (e.g., Ethernet) to another (e.g. token ring). Thus, layer 3 devices are often used to interconnect dissimilar subnetworks.
User Priority
FIG. 1 is a block diagram of a data link (e.g., Ethernet) frame 100 which includes a MAC header 102 and a data field 104. MAC header 102 includes a MAC destination address (MAC DA) field 106 and a MAC source address (MAC SA) field 108. Recently, a proposal was made to insert a new field after the MAC SA field 108. More specifically, the Institute of Electrical and Electronics Engineers (IEEE) is working on a standard, the IEEE 802.1Q draft standard, for adding information to MAC headers. In particular, the 802.1Q standard defines a tag header 110 which is inserted immediately following the MAC DA and MAC SA fields 106, 108.
The tag header 110 comprises a plurality of sub-fields, including a Tag Protocol Identifier (TPID) field 112, a user_priority field 114, a Canonical Format Indicator (CFI) field 116 and a Virtual Local Area Network Identifier (VLAN ID) field 120. The user_priority field 114 permits network devices to select a desired priority of data link frames. In particular, in an IEEE appendix, referred to as the 802.1p standard, the is IEEE has defined eight possible values of user priority (0–7), each of which is associated with a specific traffic type. The proposed user priority values and corresponding traffic types specified in the 802.1p standard are as follows.
User Priority ValueTraffic TypeDescription1Backgroundbulk transfers2Spare/Reservedn/a0Best Effortcurrent LAN traffic3Excellent Effortbest effort type of services(e.g., for an organization'smost important customers)4Controlled Loadimportant businessapplications5Video (< 100 milli-minimum jitterseconds latency andjitter)6Voice (< 10 milli-one-way transmissionseconds latency andthrough the LANjitter)7Network Controlcharacterized by a “mustget there” requirement tomaintain and support thenetwork infrastructure
An intermediate device may provide a plurality of transmission priority queues per port and, pursuant to the 802.1p standard, may assign frames to different queues of a destination port on the basis of the frame's user priority value. For example, frames with a user priority of “0” are placed in the “0” level priority queue (e.g., non-expedited traffic), whereas frames with a user priority of “3” are placed in the level “3” priority queue. Furthermore, frames stored in a higher level queue (e.g., level 3/excellent effort) are preferably forwarded before frames stored in a lower level queue (e.g., level 1/background). This is commonly referred to as Priority Queuing. Thus, by setting the contents of the user_priority field 114 to a particular value, a device may affect the speed with which the corresponding frames traverse the network.
If a particular intermediate device has less than eight priority queues per port, several of the IEEE traffic types may be combined. For example, if only three queues are present, then queue 1 may accommodate best effort, excellent effort and background traffic types, queue 2 may accommodate controlled load and video traffic types and queue 3 may accommodate voice and network control traffic types. The IEEE 802.1p standard also recognizes that intermediate devices may regenerate the user priority value of a received frame. That is, an intermediate device may forward the frame with a different user priority value (still within the range of 0–7) than the one it had when the frame was received. Nevertheless, the standard recommends leaving the user priority value unchanged.
Type of Service
FIG. 2 is a block diagram of a portion of an Internet Protocol Version 4 (IPv4) compliant IP header 200. The IP header 200 is also made up of a plurality of fields, including a type_of_service (ToS) field 202, a time to live (TTL) field 204, an IP source address (IP SA) field 206 and an IP destination address (IP DA) field 208. The ToS field 202 is intended to allow an entity to specify the particular service it wants, such as high reliability, fast delivery, accurate delivery, etc., and comprises a number of sub-fields. The sub-fields include a three bit IP precedence (IPP) field 210, three one bit flags (D, T and R) 212, 214 and 216 and two unused bits 218. By setting the various flags, a host may indicate which overall service it cares most about (i.e., Delay, Throughput and Reliability). Although the ToS field 202 was intended to allow layer 3 devices to choose between different links (e.g., a satellite link with high throughput or a leased line with low delay) depending on the service being requested, in practice, most layer 3 devices ignore the contents of the ToS field 202 altogether. Instead, protocols at the transport layer (layer 4) and higher typically negotiate and implement an acceptable level of service. Version 6 of the Internet Protocol (IPv6) similarly defines a traffic class field, which is also intended to be used for defining the type of service to be applied to the corresponding packet.
Recently, a working group of the Internet Engineering Task Force (IETF), which is an independent standards organization, has proposed replacing the ToS field 202 with a one octet differentiated services (DS) field 220. The first six bits of the DS field specify a differentiated services codepoint while the last two bits are unused. Layer 3 devices that are DS compliant apply a particular per-hop forwarding behavior to packets based on the contents of their DS fields 220. Examples of per-hop forwarding behaviors include expedited forwarding and assured forwarding. The DS field 220 is typically loaded by DS compliant intermediate devices located at the border of a DS domain, which is a set of DS compliant intermediate devices under common network administration. Thereafter, interior DS compliant devices along the path simply apply the assigned forwarding behavior to the packet.
Although layer 3 devices, like their layer 2 counterparts, typically have multiple priority queues per port or interface, layer 3 devices often apply scheduling patterns that are more sophisticated than simple Priority Queuing. For example, some layer 3 devices forward one packet from each queue in a round robin fashion. Another approach, referred to as fair queuing, simulates a byte-by-byte round robin to avoid allocating more bandwidth to sources who transmit large packets than to those who only send small packets. Another approach, called Weighted Fair Queuing (WFQ), allocates more bandwidth to specific traffic flows or sources, such as file servers, based on source IP address, Transmission Control Protocol (TCP) or User Datagram Protocol (UDP) source port, etc.
Some networking software, including the Internetwork Operating System (IOS) from Cisco Systems, Inc., support the creation access control lists or filters, which are typically used to prevent certain traffic from entering or exiting a network. In particular, certain layer 3 devices utilize access lists to control whether routed packets should be forwarded or filtered (i.e., dropped) by the device based on certain predefined criteria. When a packet is received by such a device, it is tested against each of the criteria statements of the corresponding list. If a match is found, the packet is either forwarded or dropped as provided by the list. The criteria may be source address, destination address, or upper-layer application based on their TCP/UDP port numbers. For example, an access list may allow e-mail to be forwarded but cause all Telnet traffic to be dropped. Access lists may be established for both inbound and outbound traffic and are most commonly configured at layer 3 devices located at the border of a network (i.e., gateways or firewalls) to provide security to the network.
Congestion Control
Congestion typically refers to the presence of too many packets in a subnet or a portion of a network, thereby degrading the network's performance. Congestion occurs when the network devices are unable to keep up with an increase in traffic. As described above, a layer 3 device typically has one or more priority queues associated with each interface. As packets are received, they are added to the appropriate priority queue for forwarding. Nevertheless, if packets are added to the queue faster than they can be forwarded, the queue will eventually be filled forcing the device to drop any additional packets for that queue. The dropping of packets when a queue is full is referred to as tail drop. The point at which tail drop occurs, moreover, may be configured to something less than the capacity of the queue.
Since tail dropping discards every packet over the queue limit, it often affects multiple upper layer applications simultaneously. Furthermore, many upper layer applications, such as TCP, re-send messages if no acknowledgments are received. Thus, the presence of tail dropping can cause global synchronization among upper layer applications, significantly exacerbating the congestion problem. To avoid global synchronization, some layer 3 devices use Random Early Detection (RED), which selectively drops packets when congestion first begins to appear. By dropping some packets early before the priority queue is full, RED avoids dropping large numbers of packets all at once. In particular, when a calculated average queue depth exceeds a minimum threshold, the device begins dropping packets. The rate at which packets are dropped increases linearly as a function of a probability constant. When a maximum threshold is reached, all additional packets are dropped. An extension to RED is Weighted Random Early Detection (WRED), which applies different thresholds and probability constants to packets associated with different traffic flows. Thus, WRED allows standard traffic to be dropped more frequently than premium traffic during periods of congestion.
Service Level Agreements
To interconnect dispersed computer networks, many organizations rely on the infrastructure and facilities of service providers. For example, an organization may lease a number of T1 lines to interconnect various LANs. These organizations typically enter into service level agreements with the service providers, which include one or more traffic specifiers. These traffic specifiers may place limits on the amount of resources that the subscribing organization will consume for a given charge. For example, a user may agree not to send traffic that exceeds a certain bandwidth (e.g., 1 Mb/s). Traffic entering the service provider's network is monitored (i.e., “policed”) to is ensure that it complies with the relevant traffic specifiers and is thus “in-profile”. Traffic that exceeds a traffic specifier (i.e., traffic that is “out-of-profile”) may be dealt with in a number of ways. For example, the exceeding traffic may be dropped or shaped. With shaping, the out-of-profile traffic is temporarily stored until the demand drops below the threshold. Another option is to mark the traffic as exceeding the traffic specifier, but nonetheless allow it to proceed through the network. If there is congestion, an intermediate device may drop this “marked” or down graded traffic first in an effort to relieve the congestion. Another option is to change the accounting actions for this out-of-profile traffic (i.e., charge the user a higher rate).
Allocation of Network Resources
As shown, computer networks include numerous services and resources for use in moving traffic around the network. For example, different network links, such as Fast Ethernet, Asynchronous Transfer Mode (ATM) channels, network tunnels, satellite links, etc., offer unique speed and bandwidth capabilities. Particular intermediate devices also include specific resources or services, such as number of priority queues, filter settings, availability of different queue selection strategies, congestion control algorithms, etc. Nonetheless, these types of resources or services are highly device-specific. That is, most computer networks include intermediate devices manufactured by many different vendors, employing different hardware platforms and software solutions. Even intermediate devices from the same vendor may be running different software versions and thus provide different functionality. Thus, there is no consistency of resources at each of the intermediate devices and, therefore, it is generally not possible to simply select a single set of parameters for use in configuring all of them.
In addition, the allocation of network resources and services is becoming an important issue to network administrators and service providers as greater demands are being placed on their networks. Nonetheless, at the present time, there are few if any techniques available for applying traffic management policies across a network. Instead, the allocation of network resources and services is typically achieved by manually configuring the interfaces of each intermediate device. For example, to the extent there are parameters associated with a particular queuing strategy available at a given intermediate device (e.g., queue length for tail drop and minimum, maximum and mark probability for RED), these parameters must be set device-by-device by the network administrator. This is a time consuming and error prone solution. In addition, there are few if any tools currently available to network administrators suggesting how various network resources and services might be coherently allocated in order to implement any general policies. Accordingly, the ability to allocate network services and resources to implement network-wide quality of service policies is difficult and time-consuming.